Fallback Chain — OpenAI → Anthropic

​

Policy JSON

{
  "reliability": {
    "fallback_chain": ["gpt-4o", "gpt-4o-mini", "claude-3-sonnet", "claude-3-haiku"],
    "action_on_chain_exhausted": "DENY"
  }
}

policy = {
    "reliability": {
        "fallback_chain": ["gpt-4o", "gpt-4o-mini", "claude-3-sonnet", "claude-3-haiku"],
        "action_on_chain_exhausted": "DENY",
    }
}

​

What It Does

• Routes requests to the first available model in the chain
• Falls back automatically when a model returns an error or the circuit breaker is open
• Cross-provider resilience — if OpenAI is down, requests route to Anthropic
• Denies the request only if all four models in the chain are unavailable
• Works with TealCircuit to skip models with open circuit breakers

​

Example

{
  "action": "ALLOW",
  "risk_score": 0.3,
  "reasons": ["fallback_activated"],
  "evidence": {
    "requested_model": "gpt-4o",
    "fallback_chain": ["gpt-4o", "gpt-4o-mini", "claude-3-sonnet", "claude-3-haiku"],
    "attempted": ["gpt-4o", "gpt-4o-mini"],
    "routed_to": "claude-3-sonnet",
    "fallback_depth": 2
  }
}

​

Install

# Install via CLI
tealtiger policy install fallback-chain-openai-anthropic

# Or copy the JSON above into your tealtiger.config.json

​

Customize

​

Related

• Circuit Breaker — Standard — Integrates with fallback chain to skip failing providers
• Retry Budget — 3 — Retry before falling back
• Model Allowlist — Ensure fallback models are in your allowlist
• TealCircuit Documentation — Full API reference